The precise prediction of storm surges is one of urgent issues for humankind. However, the most important information such as spatial-temporal variations of air pressure and wind field for the storm surge warning is still limited. Typhoon MAEMI landed on the southeast coast of the Korea on 12th September, 2003 was simulated. MAEMI recorded not only tremendous economic loss but also historic weather survey in Korea. In this paper the importances of the meteorological input data and the multi-nesting grid to the storm surge prediction were explored by a series of numerical experiments. All numerical experiments were compared with 11 tidal station records